PlanE Framework Optimizes Data-Tuning-Inference for Extractive LLMs

PlanE introduces a Data-Tuning-Inference planner to automate the selection of base models and instruction-tuning strategies for specific datasets. The framework reduces annotation costs by decomposing data and optimizing the construction of task-specific extractive models.